[Pgpool-general] inconsistency when using two pgpool instances in warm standby?

[Jaume Sabater]

Tatsuo Ishii wrote:

>> I'm designing a warm-standby Postgres/pgpool system accepting 
>> connections from a number of application servers.
>>
>> Of course, running a single instance of pgpool introduces a single point 
>> of failure.
> 
> To avoid the SPOF, you could use pgpool-HA or whatever Hot/Standby
> additional software to construct pgpool HA cluster.

This is the type of setup I have running (Heartbeat controlling two resources: service IP and pgpool2). Writing an article about it at the moment, will hopefully be ready in a couple of weeks.

Whether pgpool-II runs on the same machines as PostgreSQL, the same solution applies.

> Yes, I had gotten this kind of questions many times. So I talked to
> someone who are familiar with networking. Suggestions from him was:
> 
> 1) use bonding/teaming between pgpool("A" or "B") and switch. This
>    will virtualy make networking down between them never happen.
> 
> 2) we don't need to care about networking failure between db servers
>    and switch since if it goes down, each pgpool instance will do the
>    same failover anyway.
> 
> 3) any failure inside switch does not likely cause the scenario you
>    mentioned.

/agree

-- 
Jaume Sabater
http://linuxsilo.net/

"Ubi sapientas ibi libertas"